World of Science | The next generation of weather models

Dylan Reynolds is a PhD student at the SLF and researches how high-resolution weather simulations can be improved and coupled with snow models. In the following article, he explains his research and why avalanche forecasting can also benefit from faster weather models.

The steep, complicated topography of the Alps is a blessing for anyone who enjoys skiing and mountaineering. From rugged peaks, the route descends steeply into deep valleys. The views are beautiful, but the complex terrain is a nightmare for the atmospheric models that are supposed to predict the weather here. Very high-resolution models are needed to predict snowfall in the mountains. Avalanche warning services often use snow and wind information that is statistically generated from a range of different data products. This is useful, but with high-resolution weather models you could calculate the underlying processes directly instead of relying on spatial statistics.

Why isn't this done? The computing power required for such models is simply too great. A lot is being invested in the next generation of weather models, but it will probably be at least another 10 years before we have operational numerical weather forecasts on such scales. To improve the forecasts anyway, we can look for ways to simplify the complex, high-resolution models without compromising the results too much.

The first results show that the combination of HICAR and a snow drift model can simulate drifting on the scale of individual ridges and terrain edges. An important prerequisite for predicting drift snow danger using a model! We hope that HICAR will eventually be able to calculate drifting snow volumes and hazard locations on individual slopes to support the hazard assessment of the warning services.

HICAR has already been used to simulate winter precipitation over the Swiss Alps with a spatial resolution of 250m. This has worked well, in some cases better than with other methods. Winter precipitation is particularly important for the SLF's Operational Snow Hydrological Service (OSHD). A snow model with a similar resolution is operated within this framework. We are confident that HICAR can be used operationally in the coming years and combined with the OSHD model. With its special downscaling method, HICAR would represent a kind of bridge between the less high-resolution weather data from MeteoSwiss and the OSHD snow model. This model chain could simulate snow processes in the mountains better than the individual models alone have been able to do so far.

HICAR will not replace the existing weather forecasts. HICAR's shorter computing time is due to the fact that it uses a lot of data from the other models and does not have to calculate everything itself first. HICAR will not be used instead of other models, but with them. Model chains and downscaling methods such as those used in HICAR can help us to better understand the complicated processes that take place in the mountains between the atmosphere and the snow cover.